Growth mindset

A growth mindset is the belief that ability is something you build. A fixed mindset is the belief that ability is something you have. Carol Dweck named the pair in the 1990s, the 2006 book sold into the millions, and the distinction is now taught in schools, printed on staffroom posters and used in job interviews.

Then the effect was tested at scale, twice, by people who wanted to know rather than to confirm. It survived. It also came back much smaller than the posters — small enough that most of what is claimed for it in a corporate training deck is not supported by the study the deck cites.

The honest position is not "it was debunked" and it is certainly not "believe in yourself harder". It is narrower and more useful than either: a mindset is a belief about one domain, it is measurable, it moves, it predicts a little, and the intervention built on it helps a specific group of people by a specific small amount.

This page is the one to read before or after our Growth Mindset test, because what that test can and cannot tell you follows entirely from the paragraphs below.

The original scale is short and blunt. It asks whether you agree with statements like you have a certain amount of intelligence and you can't really change it. Agreement is a fixed mindset; disagreement is a growth mindset. That is the whole instrument.

Two things follow and both get lost.

It measures a belief, not an ability. A growth-mindset score says nothing about how fast you learn. It says what you think about how fast you learn. Those are different quantities and only one of them is on the questionnaire.

It is domain-specific. Blackwell and colleagues measured beliefs about intelligence, in adolescents, over a school transition. A person can hold a growth mindset about mathematics and a rigidly fixed one about their own temper, their singing voice or their ability to be on time. Treating the score as a general personality dimension — a kind of person — is the most common misreading of the entire literature, and it is exactly the error the entry on psychological essentialism describes.

This is the part a page selling a quiz is tempted to skip.

The correlation is small. Sisk and colleagues meta-analysed 273 studies covering over 365,000 students and found the average association between mindset and academic achievement was about r = 0.10. That is a real relationship and it is a weak one: mindset accounts for roughly one per cent of the variance in achievement. For comparison, prior grades predict later grades several times more strongly.

The intervention works, on a subgroup, by a little. The National Study of Learning Mindsets randomised over 12,000 ninth-graders across 65 US schools to a short online growth-mindset exercise. Yeager and colleagues found no detectable effect on higher-achieving students and a small one on lower-achieving students — on the order of a tenth of a grade point, concentrated in schools whose peer norms supported the message. The result held. It was also, by the authors' own framing, a cheap intervention with a modest targeted effect, not a transformation.

The averages across all studies are near zero. Macnamara and Burgoyne's 2023 meta-analysis of mindset interventions found the overall effect on achievement was very small and not reliably different from zero once study quality, publication bias and pre-registration were accounted for, with the clearest signals again in disadvantaged and low-achieving samples.

So: the finding narrowed. It did not vanish. What vanished is the version where a mindset poster changes an ordinary student's trajectory.

People take a growth mindset test to find out whether they are the kind of person who can change. The instrument cannot answer that, for three reasons that are worth stating plainly on a page that links to one.

The items are transparent. Almost nobody reads you can't really change how intelligent you are and fails to see which answer is the approved one. That makes the scale unusually easy to answer as the person you would like to be, which is the general problem covered in response styles, in an unusually pure form. Scores drift upward as the idea becomes better known, which is a property of the culture rather than of the people.

Self-reported mindset and behaviour under difficulty come apart. Saying ability is malleable is cheap. The behaviour the theory is about — staying with a problem after the first failure, choosing the harder assignment, asking for the feedback you will not enjoy — is not what the questionnaire samples. It samples your endorsement of a slogan.

A single number hides the domain. A middling overall score can mean moderate beliefs everywhere, or a growth mindset about work and a fixed one about relationships. The second is far more common and far more actionable, and an average erases it.

Our own Growth Mindset test has the same limitations as every other one, including the transparency problem, and it is worth saying so on the page that sends you to it. What it is genuinely good for is the domain question: which parts of your life you narrate as fixed. Read the section scores, not the headline.

Three things survive the shrinkage intact.

Fixed beliefs are real and they are local. The sentence people actually live by is rarely "intelligence is fixed". It is "I'm not a numbers person", "I'm just bad at confrontation", "I've never been able to sleep". Each one is a fixed-mindset claim about a specific domain, each one is testable, and most of them have never been tested.

Praise for process beats praise for traits. This part of the literature is older and sturdier than the intervention studies. Telling a child they are clever and telling them they worked the problem carefully produce measurably different responses to the next failure. The same holds in a performance review.

The intervention is targeted, not universal. The effect concentrates in people who are struggling and who believe the struggle is evidence of a ceiling. If that is not you, the mindset exercise is unlikely to do much, and a page that implies otherwise is selling.

Most of those sentences are about work, which is where the domain question gets concrete: the Work Style and Career Values results are a better place to find the specific belief than a mindset score is, because they ask what you actually do and want rather than what you think about change in general.

The useful move is not to take a test and adopt a label. It is to pick one sentence you say about yourself that begins I'm just not — and treat it, for four weeks, as a hypothesis rather than a description. That is the whole mechanism, and it does not require agreeing with a questionnaire. If the sentence you picked turns out to be about effort disappearing rather than ability, the relevant page is burnout, not this one.